xref: /aosp_15_r20/external/ComputeLibrary/src/core/CL/cl_kernels/common/permute.cl (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1/*
2 * Copyright (c) 2018-2021 Arm Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24#include "helpers.h"
25
26#if defined(DATA_TYPE) && defined(DEPTH_IN) && defined(P1) && defined(P2) && defined(P3) && defined(P4)
27/**Perform a permute operation on an input tensor of Shape DCHW.
28 *
29 * @attention Data type can be passed using the -DDATA_TYPE compile flag, e.g. -DDATA_TYPE=float
30 * @attention Input tensor depth should be given as a preprocessor argument using -DDEPTH_IN=size. e.g. -DDEPTH_IN=16
31 * @attention Permutation vector is passed as a preprocessor arguement using -DP1, -DP2, -DP3 and -DP4=int, e.g. -DP1=2, -DP2=1, -DP3=0 and -DP4=3.
32 *
33 * @param[in]  input_ptr                            Pointer to the source image. Supported data types: All
34 * @param[in]  input_stride_x                       Stride of the source image in X dimension (in bytes)
35 * @param[in]  input_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
36 * @param[in]  input_stride_y                       Stride of the source image in Y dimension (in bytes)
37 * @param[in]  input_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
38 * @param[in]  input_stride_z                       Stride of the source tensor in Z dimension (in bytes)
39 * @param[in]  input_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
40 * @param[in]  input_offset_first_element_in_bytes  The offset of the first element in the source image
41 * @param[out] output_ptr                           Pointer to the destination image. Supported data types: same as @p input_ptr
42 * @param[in]  output_stride_x                      Stride of the destination image in X dimension (in bytes)
43 * @param[in]  output_step_x                        output_stride_x * number of elements along X processed per workitem(in bytes)
44 * @param[in]  output_stride_y                      Stride of the destination image in Y dimension (in bytes)
45 * @param[in]  output_step_y                        output_stride_y * number of elements along Y processed per workitem(in bytes)
46 * @param[in]  output_stride_z                      Stride of the source tensor in Z dimension (in bytes)
47 * @param[in]  output_step_z                        output_stride_z * number of elements along Z processed per workitem(in bytes)
48 * @param[in]  output_stride_w                      Stride of the source tensor in W dimension (in bytes)
49 * @param[in]  output_step_w                        output_stride_w * number of elements along W processed per workitem(in bytes)
50 * @param[in]  output_offset_first_element_in_bytes The offset of the first element in the destination image
51 */
52__kernel void permute(TENSOR4D_DECLARATION(input),
53                      TENSOR4D_DECLARATION(output))
54
55{
56    Tensor4D in  = CONVERT_TO_TENSOR4D_STRUCT(input, DEPTH_IN);
57    Tensor4D out = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(output, 0);
58
59    int out_index[4] = { 0 };
60    int in_index[4]  = { 0 };
61
62    in_index[0] = get_global_id(0);            // W
63    in_index[1] = get_global_id(1);            // H
64    in_index[2] = get_global_id(2) % DEPTH_IN; // C
65    in_index[3] = get_global_id(2) / DEPTH_IN; // B
66
67    out_index[0] = in_index[P1];
68    out_index[1] = in_index[P2];
69    out_index[2] = in_index[P3];
70    out_index[3] = in_index[P4];
71
72    *((__global DATA_TYPE *)tensor4D_offset(&out, out_index[0], out_index[1], out_index[2], out_index[3])) = *((__global DATA_TYPE *)in.ptr);
73}
74#endif // defined(DATA_TYPE) && defined(DEPTH_IN) && defined(P1) && defined(P2) && defined(P3) && defined(P4)
75